Você está na página 1de 24

. ..

...Ir?c,,<

:
:,,0

:
... .,.$..

,.,.. ,,.,. ,.,.! ,,, ..,,. =,,,., !.,8,,,,,.,,,,,, s,.,.,,.

THE SOCIETY OF NAVAL ARCHITECTS AND MARINE ENGINEERS 07306 601P.m!l. Ave...,uite00, S 4 Jersey City, New Jersey

S,...,, ..,,...4 ..,., ,,,,,,,,.., .,,,,.,., oa. ber G.,,7 5

@ +,

A Structural System Reliability Care Study of an Eight-leg Steel Jacket Offshore Production Platform
H. Nordal, Statoil, .stavanger, Norway C.A. Cornell, Stanford University, Palo Alto, California A. Karamchandani, Stanford University, Stanford, CA
ABSTRACT ness) The ability of the structure to withsts.nd damage The eight-lea, form wave reliability steel-frame piling) of an idealized offshore under platextreme using anaresisvaxiwas measured (structural Robustthe

by comparing failure

increase?. system for tbe damaged that The found framing

probability relative structure. were found to to

(excluding loads

structure intact

has been system loads

estimated

for the

structural lysis. tances ables, were Wave were

reliability and menber as rando~

alternatives sensitive

he comparably Calculated probability orders

to damage.

treated

increase

in failure one and two

and reliability under

calculations a variety order of reliabi-

is between

performed

of magnitude.

conditions

using The

first

lity methods. include

issues versus

examined K-bracing bents, on in

NOMENCLATURE

X-bracing

Ax the principal effects system load-carrying the D d of wave load variability

cross damaged
outer inner modules force mean Gulf

sectional state
diameter diameter

area

redundancy, conditions removing

hypothetical that selected were frame simula-

of member of member

damaged ted by
members,

di E

and the s.dequacy of bracing frame to transfer bents. analyses failure loads

of elasticity or force factor

horizontal between static brittle the

F f Y GM I k NS f F. B L o

Deterministic resulted in

yield

stress

push-over system

with

no post Using

of Mexico state factor

first member pxGpOsed measures

failure

capacity.

intact buckling North

probabi 1istic the x-braced an order

redundancy frame w-as fcu~d of magnitude than the (as

to be alnost Letter K-braced measured


framing increase decreased.

Sea of failure

(more =edundant) frame. in this


alternatives as wzve load

probability resistance safety buckling post index

The

redundancy

report)
was

of both
found to

variability

length.

failure fraction of ur, failed capacity factor

correlation

193

l---

7NTROEUCTION

Ttw Tn the recent years there


0?

s+-ucture using
by

has

previously

been

has

studied
awIo Lch

a ueteswi r)istic (2) . It

eveiopmenc bee,, a ccr. tir.tmus r?. system

I.loyd and Clawson possible

reliability

methcds

carakle

cf

was

therefore

to cor>cer:trate aspects suck.

an. z.lyzing corxplex structural In the present UFOP study,

systems. based

c: ffcrt

on probabil istic models

a method

as uncertainty
ronment and. the and

for the envi(member on tk.c

the member-replacement in Guenard study

technique

structure in

as described applied leg,

(1) has been of an eiqht-

capacities) systcm

particular analysis.

in a case

reliability

steel-jacket The structure

platform. is shown in Fig. in for the thus does or of more A brief analysis description is given of the beluow. >. of structural methods method
ANALYSIS METHOD

1. It is an i.cfealized structure the sense that it is designed only, and

operation not

1 phase

include

any added steel. be kept

fabrication importance if an the

insta?.lation this should

The

general

discussion

in mind

systems is given

reliability

analysis

attempt. is made results of the

to generalize study. Lateral

in Karamchandani analysis

(3) . The program jn

forces primariand by

reliability FAILUR Guenard

computer in detail

are transferred

to the piles diagonals

is described (1) . The

ly by the broadside the x-braces bents. X-braced een in the

analysis

estimates

four vertical horizontal loads In our betwstudy of

the probability to extreme defined the wave

of systew% ?ai>. ure due loads. Failure is of

At each panels

level,

transfer bents. the

for each

individual

nember function ?.nemher

the vertical est imated when

structure

by a failure state.

we have

reliability to

describing

a limit

the structure extreme wave

subjected we have

fails if a limit state is reached. The limit state or failure function .sed for the truss members of tbe that to

loads.

focused example jacket is the condition force is equal When

on the behaviour bents Tbe

of the

four vertical K-braced) perforits elastic resistance. stiffness force a member fails its drops to zero, a confrac(The forcefor tbe is . the axial member

(first X-braced is limited

then

study

to their wave

mance namely scuth.

vr.der the worst a broadside Next we have and

direction, and the rvmber becomes

wave

from ti,e tbe wave stant, equal


to a postfail,, ~e

varied

load variability

investigated

its tion of its maximum of the capacity. details semi-brittle assumed stricture

effect on the reliability of the two structural systems. Further we have evaluated OUSIY the consequences damage, and of exogen finally

de formation truss

bebaviour of the

caused

members

examined

the behaviour that

of the horitransfer bents. loads the

discussed

later. )

zon.+.. X-braces 1 between wave

the vertical

For

The ure of the stiffness tion

?c.T2?. rriterion system

for

fail-

direction

analyzed are almcst

the horizon-

ib z major

1.OSS in

tal X-panels the intact

unloaded

in

as measured

by the deflec-

structure.

at the centex

of the top of the

structure.

Pra. +.ic?,l.ly he analysis t path was or st??uence of stopped when tb. e

of a particular membe]- failures

194

!
,6
15.25 4.6

:i m
i LEVEL PLk. ,
-x
&16.9+12.2+16.9+ !-12.2-l-12.2-i-

,,..4

Row 1

,-

Fig.

General Son

view

of exmple
in

structure.
meter.

Froxa Lloyd

and Claw-

(2) . Dimensions
sequence

probability

of

the

stopped after had can be of its 4

paths failure lf,

are

ignored,

the

true be

system larger) . all are the not path) , of true

decreasing; to 6 members failed. defined menbers. failures ure

this was in the

usually structure system

probability specific

will

for any

path,

Failure in terms Each which

of the

metier

failures

in the path

of failures

considered then

(i.e. an incomplete

sequerce leads

of member to system path; failthe

the probability will

of occurrence than the

the path value

be larger

is called

a failure

and a system all


in of such an

failure
incomplete UPP er

analysis
paths on In could of , the tbe

probability
probability will not occur. included

of system
any Ii in such all the

failure
failure

is the
path

including
will result

bound

possible analysis, is of a

paths then

are

probability extreme he any and the one the case

system lower of

iailure. bound occurrence

, the

t!le result on the

obtained

lower

bod failure

probability particular !ipper Lound

Frrh:bi.lity since some

system

cor.plete COUIC be

sequence the

(i.e.

pctential

failcre

195

probability the intact

thet structure

?.r.y

member (i .e.

fails in each

in

with

>,:r. Gne member t state with

failed

or a r.,en!bers

damaged is

two

failed

path

only

the

first

member fur< her


cf

failure fc:iures

(i.c. , member member

i, and

another

considered are
t ions ignored) made,

acci ali
. the Because

j ) . Subseque.r,t. in thi E mcst 1ikei:- dauagecl next. The

aFproxiruaused

iailures state

lower

bound

?.re ?r.vestigated continwss focused

cannot, except pure

strictly under

spe.?.king, be proved conditions, behzviour. e .g.

procedure state state the

until

the c?anaged

special

being

on is a collapse state in which

elastic-plastic

(i.e. a damaged

systen

is ccmsidered sequence of

to be failures state This is the is

For

most

realistic number

structures, of possible techthe

failed) . Tbe leading most

there

are

a large

to this

damaged

failure
nique

paths.

Hence

a search
to

likely

failure because feature This

path.

has

to be used
paths (i.e.

identify which occur-

guaranteed look-back algorithm. continued

of the ur.iqe in this can be at the next state) failure

important have rence) a high . In

paths of

probability Guenard

procedure

(1) , a branch-

(i.e. , looking tG OCCUI other

and-bound most

technique

to identify path

tbe

most

likely

damaged

important

failure

has been the the will occur

to generate paths.

important

described. first step

In this

technique

in to investigate that a failure

probability in the member that

The
has form for a

computer to

program automatically

FAILUR persearch III the

intact i, the

structure. For probabi 1 ity f ai 1s, P the

each

opticm

such the

a brac-and-bound important paths. being many of

the

member

most

(member intact this

i fai 1s in the structure)

idealized

structure because

studied, tbe critical

is computed.

Note; that a

however,

is also state

the probability with member i

member critical failure number tigated

(such as the X-braces bents) have very

in the

damaged failed

similar

is reached.

Let member the largest

probabilities, of damaged is very

the total to be inves-

i, be tbe member probability damaged of

with

states Iarqe

failure. associated

hence , the with member

(i.e. many with simi-

state

alternative

damaged

states

i, failed
to occur shifts next

is tbe most damaged

likely Focus state. ow The

lar probabilities exist) memory and the

of occurrence time
large

state.

computaticm is too

(and
to

to this damaged

requiremmtl

step

is to investigate of subseqet that

the proThe

be practically puter. Therefore

feasible i.n this paths

on a minicomstudy were the

bability

Iailres.

probability state with

a subsequent

damaged

Fctential fied using the paths

failure

identiz.lthcmqb,

nrar,ber i, and is P
in the

ICZCK21 searches. identified using

failed

occurs
i, fails

such manto be expeusu

(member

al searches j with the most dience ally surviving

are not guaranteed paths, past

intact fails member

structure

ar.d member structure this

important

in a damaged

irxlirates that

this will

i, failed) . Once ca Icclated the focus

be the case.

In tact

in many only one

has been members,

for all shifts

cases,

in the manual failure

search, was

to the state .

important

path

ide[>i-i is the

crrentl:, most This ccmld

likely

damaged

be either

a dam+.$ed state

fiei. AC, c>. !.srus:e< below, this usually sufficient to deccribe

196

-.

system

belmvic,ur sense.

in the

syskem

reli-

ability

The presented tlees

.esults in the

cf tl.. scarcl.cs, a., e , fern cf friiluye the nodes of the enervating

In Lhcse trees, damaqed a,d

represent structure

states

the branches

from a node

represent

n,en;ber failures damaqeci strucin these in Fig. 2.

in the corresponding tures. failure The

otati.cn used is explained

trees

In addition
tic analysis a set

to of

the

probabiliswith ically was

anaiys~s

member equal carried pare tural and to

resistances to the cmt. with model the of mean

determinist resistance

The:: \Jexe used


more Lloyd with advanced and the For probability exceed

to comstruc-

2
WIAC1
Srmlcla,

,s7

---PWJmO, ,cm, ,.,,,CMAC,T, L,TV m .WSER ,,w,. SF,Xc,,c+o,. S,RwruR, ,..,..,*. ,4,0 ,,,.,,.
S,WC711RE ,,ACT ,s
PmmAwLmv ,0.6, TlwwAcmv m OF M,-R ,4nwm EXcz,m. wnm

14 3.,7

Fig. (2) ,

Failure

tree

notation

Clawson probabilistic example, that simply

compare

space
one tbe the

truss

model

representation

of

analysis can

performed. the will

the

structure and

consisting 67 nodes.

of 216 truss The foudat.ion

compute loads

elements is rigid.

random

deterministic

system

resistance.

The MODELLINC OF THE CASE STUDY according

structure to the API

was

designed sing

guidelines

all X-braces. Structural


Model

Dimensions are given were sized

of the veri Fig. simply 3. accoras

tical

X-bents

Some members The as simple order listic results used structural as could model was made ln ding

to general member

requirements slenderness wall

such

be justified.

maximum

, minimum thickness.

to concentrate aspects of the

or, rAe probabistudy, the (2) were the

diameter, Examples horizontal o f these bee

and minimum

of such mer.hers are the x-panels . The dimensions members would normally temporary to the operaHowno such have

of Lloyd

and Clawso

.s guidance

in identifying behaviour Lloyd . and

dGrninant structural In their Clawscfi report 61:Y that act like

dictated

by some prior

study, the

design tional evex,

condition phase

structure and

to glohfurther to franc to the ir.

of the

structure. study

a truss,

for this was

example explicitly

the main i.n the wave

contribti.o structure

condition

considered.

action direct

is due

loads

on the members zone.

The AFI used given testing value. tG design in Table

rec Gnmende? the

kfactors are

and below quence

the wave

As a consethat ylokal

structure

it was

anticipated

1. Based , more

on recent xeulistic by

beha.viour could tr~ss eiement

be model led by s. if the wane lc,acs

expedience of k 1s were at Exxon

model

recaunended

arc treated loads. lhese

as concentrate.t assumptions

~oint to a

engineers tl!e mea

far calcla.tinq of the jacket

lead

resis+.ace

197

Table

I. Table API

of buckling values

lengths, i , a)xl reiucticn factors k. e..,< assumed mean values. stud:,

design

Member type

Eesiqc k,i

cscd in this to predict

mean capacity

k ,L X-brace K-brace Diagonal Leg Horizontal to leg brace leg 0.7 ,L 0.5 ,L brace
0.9 ,E12 0.5 .!.12

0.8 0.8 1.0

,t ,L , i

0.7 0.5 0.5

, i ,! ,1

* L is for a brace

the

length

along

the brace

from

leg to leg.

members. given factor the

These

k-factors 1. Note

are also that the kthat

cal

in this

respect.

in Table used

for X-braces members respect

implies are fully

The shown cribe

fcrce-deformation 5 was applied

diagram to desIn this force

compression with

in Fig.

supported out

to buckling

the member

properties.

of plane

by the

intersecting

semi-brittle

model,

the member

tension

member.

~.ncreases elastically capacity or resistance.

to the member After failure, In the its fail-

As part ture wzs X-braces bents Fig.

of this

study

the

struc-

i .e. if the axial element

deformation beyond force

redesigned replaced

with by Ks.

vertical The vertical in by

is increased

ure value, drops failed

the element

abruptly

of this

structure

is shown sized

to a fraction, capacity. For value

q , of its unthis ?.pplicaticr>

4. The K-braces the values

were

mztcbing force checks

of only

the axial unity terms Other been total in

a deterministic

of q = 0.4 was

term with

in the API-guideline the corresponding X-braced design. have

used for members


sion and In other
tension the

failing in compresfailure. ductile


maintaining abrupt

n = 1.0 fur tension words


failure load,

the oxiginal matching used, unity level. matching had

we assumed
behaviour, and an

criteria

might weight,

e.g. , equal check The

equal

failure

drop

or equal combination

reliability of the chosen values be

to 40 % capacity

when failing in comin the light

idity of these pression. The Va!. assumptions of the is discussed results.

rule

and the k-factor influence, as

a profound

will

analysis

seen below. Tn a probal,ilistic The framing twice one elevation not of the hcri zontal the batter This is of member mean properties context the rarxium

arf? flescxibed by z a coefficier.t oi

W*S

changed,

resistance,

the members source

in the 1:-brace is thus of the Xs. of difficulty may i.n the comEot be typi-

vari ?.tie (COV) and distl-ibution. des~qn

a probability to tfe from cedes, the

the batter

In contrast taken

capacities some

parison.

The K-braces

wb. ich. include

conservatism,

198

LG9

LG2

LG ,

Fig.

La4

LG3

LG2

LO,

Fig.

_..__ i!i!i I!!!i


z, ~B3

LG.

*Z1
822 .%3

z,

XB2

HZ2

x.,

HZ4 ml XB2 X63 xl IG2 IC3 ~

61 (24) 51 (20) ,1 116 ,6 {14) 51 (2. ) 46 {18) 36 (14) 100 (39.25) 100 (39.251 100 (39.25) 100 (39.25)

0.95 0.95 0.95 0.95 1.27 1.27 1.59 4.76 3.49 3.49 1.91

(0.315) (0.375) (0.375) [0.375) (0.500) (0.500) [0.625) {1.875) (1.315) [1.375) [0.750)

Hz,

Dimensions

of

X-braced

vertical

bents

-E,

D,M16TEP. (inch,

Wi.cnw,,s

w,

GROUP

i m

in e

(in.h)

B3 H=

ml

KB2

14Z2

KS,

Hz2 Ez3 X.4 ml =2 KB3 xl X2 X3 Z4

61 (24) 5, (,0, 41 [16) 36 (14) 61 (24) 56 (22) 41 [161 100 (39.25) 100 (39.251 100 ,39.251 100 ,39.25)

2.86 1.91 1.91 0.95 1.27 1.27 1.27 4.76 3.49 3.49 1.91

HZ1

(1.125, (0.750) (0.,50) (0.375) (0.500) [0.500, (0.500, ,1. B75) (1.375) (1.375, (0.,50, ,

Dimensions

of K-braced

vertical

bents

mean the

resistances real

are best

estimates

of duced

A 15 % reduction in the resistances

has been

intro-

capacity.

[the O .85

factor) The AISC formula given helo~ was used memb-

to compensate induced capacity

for the neglecby frame was action. as

ted moments The tension

for the capacity ers the . Note code that

of compression for our factor

calculated

application was removed. ~=0.85f yx A

safety

which R.c = 0,85 fyAx(l-~ 2CC2 2 2Tr J! r Y

is conservative factor.

due

to the

15 %

reduction

In general exanple ratjcs

the r,en,bers in the low slenderness

cc=

strx; ct, ure had

res,:.lti~,q i,,. iew rec L~cti cr. in due to

.Lz7 ,
4 199

the compres?i. on capacities buckling.

a---

Fontx --------R
I---- ---CUXILI!(, -11

In scaling should

using of the in

this fixed mind realistic tti realj

load load the

mc~e~. , pattern, danqf. r

j. .e. one of botb

keep

=M1-mRUnLEIO$,llt

scaling with.

beyond respect waves dud

values, stically

I ,n ~
BRITTLE( , -0) DEFORMATIM

Occurvelocity

ring fields

and with

associated respect due to to the with Tn our is sea

changed surface wave actual study, (50 height

structural effects. should be In

geometry particular

Fig.

Member

force

deformation

mcdel

compared

the example 15.25

The member
assumed with to be a coefficient

resistances

were

deck the feet) quently 30.5 will because in

elevation. deck elevation mean wave ( 100

Iognormm.lly

distributed, (COV) and 13

meters Conse-

of variation capacity capacities. work

above for meters be

level.

of 10% for the % for the COVs Moses

tension

heights feet) the

above

about

compression upon

The by

structure analysed

are based (4) .

presented

unconservatively we load neglect when of the tbe this

sudden deck is

increase hit. is

wave

Environmental

and

Structural

Load

(The under

importance current

assumption )

investigation.

Gravity acting the

and environmental structure

loaas were in F AZ is height, a function of the wave COV zre are and

on the jacket program each One

computer patterns

model led as two being scaled by a is

consequently

the mean, of FLk

fixed random used loads

probakili~y all

distribution given

variable.

load pattern

indirectly

h,her.these

for lateral and one

environmental dead

(wa*-e) load,

known loads

for the wave are assumed

height. here

The wave a

for vertical load

to follow Moses

Guenard used

(1) . The wave on the

pattern

lognormal discussed addition describe the

distribution, there,

(4) . AS in

is based cm the

set of forces due to the design


nodal truss

F Lk also

includes, to

acting

structure

to 11, a random the uncertainty (base shear) data

factor

19.2 meter wave.


forces model This to 1 ing

(63 foot) , 100-year


pattern be of consistent the structure. is input with as tbe

in predicting given the wave

forces

height. values

Background used

for the cOV in Moses sets (4) of F!. L

car be found

and Anclerson et al
Wave ture wave are loads acting drag to can be base on the strucThe F, for by: our

(5) . Two one

are used ting case Sea case COV

in thiz

study,

represen(GM) a Worth GM

basically H,

c?ominated. shear,

a typical (our base (WSI case, is also case

Gulf

of Mexico

height,

case) , and one see Table

transformation application

therefore

11. The

approximated

referred

to as the higb1ow-COV cOV is due

as opposed

to the

F=

CHX=F

~9 (H/19)x

= F19 x FL!

case

of the NS.

Thi s higher

in part with x = 2.0 tc 2.2, C a constant, shear wave associated with F19 the and

to tbe hurriczne cases are

risk

in the to have

GM. AS both the same

scaled

is the base 19.2 meter

100-year

Wa%-e, (O.99 fracis

pattern factor

of fcrcez,

tile) , the IJean of the NS case higher than tbe GW case due

F it iS a scale

dependent

cm H.

to the

200

lohex

COV of the

former

case

cases

include

additional

steel

ar. d.

framing The values lifetime values, gib.en in Table i.e. 11 are mects represer,tir. the g induced in the by (20 Tinis xbraced Rents

aictated

by design prior

req!,~.re -

for phases

to CIperation,

and will that

thus have

membe-f capac~ties wi.t.h loads. Vert~

n.k. ml CC7? 1. of the the extreme year) wave

forces

are more

unbalanced,

occurriug

respect

to environmental versus K-braced

lifetime in the

of the failure

structure.

results discussed

probabilities being lifetim,c

irl this

study

i the

Intact

Structure.

probabilities. X-bracedvertical base case studied with was


bents.

Table

ZI, V7ave load

scaling

factor

The

FLL . Means and coefficients of variation.

tbe

intact

structure bents

X-braced

vertical type structural was were

in a Gulf

of Mexico

I
Gulf North O* Nexico, Sea, NS GM
I

Mean

Cov

environment. behaviour obsezved.

The general ab~ve inner

discussed The two

bents

0.75
0.85
I

0.37

irtually bent were

equally also

loaded,

and the end loaded. frcminq

0.23

significantly

In addition Included a factor mately in the forces, F63, is also system forces another stiffness values DISCUSS1ON OF RESULTS was

the horizontal stiff enough

to transfer to

of 1.6 to account due

approxi-

effectively in case loss. of

from one bent failure and

for loads

to miscellaneous etc) .

appurtenances

(conductors,

Because

of tbe

low

of slenderness reduction

ratios due to

(k9,/r),

the capacity buckling

effects was

in the

compression together induced of by

General Intact

Structural Structure

Behaviour

of

members with

smal 1. However, loads

the cmnpre~sion load

the gravity When wave load subject (south behaves parallel to a broadside , the as four ?.Ey asymmeby this ties

in both tbe

members

the X s, it makes members to fail

compression

direction) basically bents.

first.

structure identical

In the deteru.inistir uniformity lead in load

analysis

try in the load the horizontal The capacity sustain lateral

are distributed frames

and capacisystem failof for

(X-braced.) . to hy in is dictated

to a brittle

of the load

structure

ure. If a post-buckling only the the 40 % ( ii= 0.4)

capacity

is assumed

the capacity tbe vertical bents

of the bents.

framing

system

semi-brittle structure

compression sustain following

members, tbe failand Even

The ertical wi tk reSEl As

cannot

are almost

identical

force

redistribution

rtispect to size til:g i.n s well will viuu~ utmn

and dimension:, balanced later

ure of first collapses with

compression load

rember ir. crease. member

structure. this heha-. in mind

without

be discussed ~.s important elaiuating

a ductile

compression

to keep

assumption tic system larger load . than


For

( q = 1 .0) , the determinisfailure l.cad is only 5 %

tb.e results

,>b Atainccl

11, cbe prc]mbil istic zeai tic is structures

analysis . Nerc wi.1 i*,r:,:u;y :.

.:.= f1r8t-rcemter
the X-brzcefl

fzilure
the

strtictxre

201

sicor=~,c~ answes . these

is

somewhere

betwceu

shown. failure

This

is the most

probable ir,dex 6 = failure ir

limits,
a gradual with

a more
model

arccrate
wculd

cmnpredict

pat:. w~,tb safety with

pressiun
only force

member

4.4. 2he patkj starts the most likely

reductior

i.nmember
Tbe

tc fail member cnd involves

in the failure X-

increased
is

def onnation. the system, has

intact

structure

consequence

that

in all members braces at the

of the vertical second level bay.

very dancy

little,

if any,

overload

redun-

as measured

by post-firstThe other are incomplete safety branches failure index in the trf.e paths. They to

nember-failure

capacity.

In the q = O. 4 deterministic analysis occurred city of the base case, failure capa-

all have

comparable failure branches in

that of the most path. were Tbe search when

probable in these the

at an ultimate

lateral

of 38300

kti (860 CJ kips) . For to follow, exceeding is 1.0 x index the

stopped index

increase one step

comparison probability

of results of che

safety ther

by going

iur In the

load

into the tree was structure

small.

this deterministic

capacity

X-bracea after had

this occurred same bay observation is that system the failure in the level

10-5
6

and corresponding This gives

safety a Reserve

4 members

at the

= 4.28.

failed. the

A general tree to

Strength ratio

Factor

(REF) defined load at to design

as the

from

failure

of environmental (undamaged)

mechanisms envicomare failure

leading

collapse romnental paring

of all members

load of 3.3 to 3.45, with the REF = 3.5

vertical levels.

X-braces

at one of the bay

well

reported based

by Lloyd

and Clawson

(2) The tree in Fig. search was 6 shows why

on a more

advanced

structural

analysis.

the automatic computer

option

in the for

program

ineffective

In the probabilistic analysis account performed here,

systems we take of lower resist-

this matic

structure. search

Following

the auto-

criteria

of searching

of the probability than average

or higher ances

member

all

for the node with the lowest 6 , s.node~n in the tree with SafetY lower prior than to the 4.4 will final be develin the In parof the the dif and

in addition

to for

the the

load

vari-

index oped most

ability. case load tively


oped Fig.

However, with the large


The

analysis COV is
tree

node

environmental reladevel in 2 failure case

probable this paths

failure leads

path.

COV, the
small.

resistance

ticular failure

to that may

developed with only

includes slightly sequence

for the base 6 with

is shown

same members ferent nearly metier the

the notation numbers

from Fiq. 7.

fai lure safety

and element Eased ysis,

from Fig. preliminary

same

?.nde::.It can error is

on an earlier only 40 of the

analwere

be shown introduced system

that only

a small

216 members

in the calculation index

of the one

formally

treated

in the probabilistic members have

safety

by ignoring

stitidy,the remaining negligible chance

of these

failure

paths.

of failing. The results of the pxobabiii.scase [case column in are

Only representing

one

complete

failure co? lapses

path is

t.~c amlysis summarized

of the base in Table III,

structcro

X-I-GM) . Second

vertical

.,.

r\

!., ., ,.

.. . .

Fig.

Failure

tree,

base

~ase

~-l_G)l

203

BENT A

Fig

7 safety

x-braced

I!i!!i I!!!
147 153 ,96 1s6 68 101 m ,09

4n

43

,1

61

BENT B

BENT c

6ENT D

vertical

bents.

Member

numbering.

the table

gives

the

index of member

(and

load,

govern

the any-first-memberThe system to

corresponding of the (MLTF) . This


found when the intact the

probability is the lowest

failure)

failure safety

probability. index

Most-Likely-To-Fail
comparing structure. safety index

is 4.2 corresponding of 1 .3x10-5. of magnitude that a

index
in column

a failure This

probability an order

all members Third

is about

less than member

tbe probability fail in the the will

gives from ures. that

calculated member fail-

will

structure. likelihood given that 10

the union

of first

Probabilistically, that the system fails

It represents at least in tbe one

the probability (i.e. any) might In tbe of tbe fail in

fail

a member % implying dancy ic

is therefore a kind

about

members tbe

system

in effect

of redun-

intact

structure.

fourth

Pot detected

in the determinist-

column, path

tbe most index fifth

probable

failure and the the of

analysis.

safety the

is given, column

finally system

gives

The base reanalyzed representing Tbe results with

case wave

structure

was

safety

index

representing

load variability Sea environment. as case the iower sys-

probability the entire

of failure structural

or collapse system.

a North are

summarized III. sees

X-I-NS The MLTF-member case is tbe member second in the base in compresbay of the 1]1

in Table one

Comparing first the

two cases, probabiliti~ tem failuIe

loaded level

of first member in the NS case. the mean that the loacs

and

sion at the mosr heavily

Recall are design cas-

loaiied inner the union it was

bent.

however scaled

that such

calculating member union

cf any noted

first that a

100 year

failure, of only

wave

height

is the

same

in both C@V,

the

6-8 most

likely in

es. Because mean scaled

C! its lower load

the in

members the

of the vertical structure sane


gives

x-braces

is 13 3 higher

intact the

approxias the union that

the NS case. of the


Fiore

The hiqhcr
COV and tbe

6 is a result
hiqh R.F.F.

mately

results This

lower

of all members. these loaded

ir.dicates al 1 are

important

for our purpcse

we

see

f m.! r::embers which in compression

a larger of any

rilffc?rence between

the unicr

by the wave

first mernher B and t.t:esystem

204

Table

111

lable of result ?.. Embers and corresponding

are

safety

index,

unclar.peu,

prcb,mhil ities

<>? failure , clamped .

Case

MLTF

h!

or. c,f zny

f:ost likely failure path

System

1st failure

failure

X-1-GM

3.96 (3.7x10 -5) + 5.10 11.oxlo -7,

3.71 (11 X1O-5)

4.40 (5.4X10-?

4 .20 (1.3x10 -5)

X-1-NS

!.97(3.

OX1O-7) -3,

5.9 FJ(1.OX1O-9)

5.94 (1.1X1O-6)

K- 1-GM

3.06 (1.1x10 -3)

2.87 (2. OX1O

3. I7(O.8X1O-3)

3. O3(I.3X1O-3)

K-1-NS

3 .84 (1.6x10 -5)

3.75 (8.8x10

-5)

4. O9(5.4X1O-6)

3.95 (3.9X1O-5}

X-D-GM (A)

3.15 (8.2x10 -4)

3. I1(9.7X1O-4)

3 .52 (2.2x10 -4)

3.50 (2.3x 10-4)

X-D-GM(B)

2.71 (3.4x10 -3)

2.71 [3.4x10

-3)

3.4 Z (3.2X10-4)

3.4 O(3.4X1O-4)

X-D-GM (C)

2.26 (1 .2x10-2)

2.25 (I .2x10-2)

3.20 (6.9x10 -4)

3.19 (7.0:< 10-4)

K-D-GM

1.52 (6.4x10 -2)

1.52 [6 .4x10-2)

1.64 (5.1xIO-2)

I .62 (5.3x10 -2)

(or failure

probabilities) words the

in the NS systmTlower-loadthis large nmdel terms

case . In other ef feet COV

.Wwoximate analysis of this structure via this simple ideal can be useful 05 faciii.sting . For our n = 1 particularly z ncre in

is larger NS. We

in the shall

case,

discuss

general we

POint

further
For both

below.
the GM and NS analys-

perspective interpret

application

(where n is the nwin the figures to

ber of components) represent the

es, as well case,

as for the deterministic failure same bay X-bracee mode tends to

system the

safety,

and n = The

a system at the

4 to represent components

system

safety. are

develop

level bents.

in all Due ~0 ad due

for or

structure

four vertical this


to

localization
prciousiy

vertically
discussed

now (two-member) X-braces. Consic?er first the characteristics of this eqcj. alent component. using the tenir, the At

the

load-

capacity

niforrnity,

tb. is system ideal

can

s$on

capacity

(t.c) of a member reference,

he ccr,psxeci to a simple lel system brittle consisting

paralsemi -is

the X-brace cmbpression

as a nit capacity

of four

is O .53 t.c. ner,ber,

components

. Such (1)

a system

faj.lure of the compression tt,e capacity trwether, of both

~j. sccssed

in Guenard in Figs.

Ej. resu:.ts s il .

members

acti~,q
capacity

zze presentea

8 thxuu~h

i.e. , the ultimate

205

1
,%.

6 -

5,.// 4 -// .. .. ..-_ .. -.__ -== --------

.=,. O

__a

~=o.,, .. .

._ ._.

___. ------

. :020_ ____ 6,, 6

3 Fig. 6 Results system


(PR=O)

;:j

V=,.25-0. O 1

cf ideal analysis

parallel

Fig.

Results system
(PR=O .50)

of

ideal

pax?.llel

ar.alysis

P. 7 SF=Z.O .

%. 8

SF=,,., 6 n - _- ..

._. __ ._. .. El
+
,,.0.0 i ,,.0.5

F=! _Ct$.o.,o

---

_ _-

__

__

SF=I.57

SF.1.4,

Fig.

10

Influence

of SF on the

Fig.

rc~undanc y o f the brittle system R SF ideal parallel

E
- ___ -- __________ 234 11 Influence redundancy brittle system
mean

Cb=o. ,,

- ___

C!+. .m o

6.

of CVP o f the

on the

ideal

parallel

(CVR=O. 10) correlation:, between factor = , L member with

(SF=2 .0)

resistances resistance and

- safety L load

R common

CVR, - coefficient
vi?riabies n - number of

of vari~tion

common

to all resistance

components

i:~ parta13.el.

206

of the X-brace 1.86 t. c. After

is: 2x(0.93 failure

t... ) = of both incr,, oi the tensitiu

between member

B ~ (systen,) ad B ~ fa~. u:=)

(any first by B a

Parl-.?.ps divided system For qiven

IJcrs the residual X- fr?Ce iS

strength

is a B-Iue2sure Gf the c~ndi.tioal reliability first member ly determinate of the

(with the ductile

rxsr.tber d the a),

60 % reducticr, in t}:. capacity] : t.c. !1,s

failure. syster,

a statical-

cc,mpression member (0.4X0.93+1.0) pest-failure

B ~= B ~ and zero. In the

t.c. = 1.37 capacity

the difference

is alays

as a fraction n , is thus: the load

of the ultimate 1.37/1.86 Cov

capacity, Further,

for a system with OppOsite extreme, extremely high redundancy @ ~ would be much larcjer than 6a and the differequal is to

= 0.74.

is 0.37

and 0.23

(for GM and NS, is found and

ence would 6. almost itself,

be approximately i.e. , the system given

respectively) from

, the net COV =

(0.37 )2+ {0.13)2 =

(0.39)2

as reliable

a member

(0.23)2+(0.13)2 effective

(0.26)2

and the

failure by this would


B~

as it is before . (Normalized tbe redundancy measure for

correlation

between is about The

colupc,lellt safety 0.89 and 0.67,

margins

perfectly,! be unity. )

redundant

system

respectively.

effective

correlation

is measured variability
the With the results from Table for become i.e. In ratios systems, of parallel , conas low 111 the

here by the ratio to total

of load

variability.

(normalized) and the

differences X-1-NS 0.12 and case, 0.16

X-1-GM

PC = (0.3712/
~

((0..37)*

(0.13)2)

Respectively: Values trast, cm from as with 0.5 a the

O to Fig. for small 8,

1 scale. we see

Unfortur.ately range study, cover does, Fig. considered Figs. our

the parameker

high even

parallel number

in Guenard *s case 11, does of the not

8 through Study show that

members,

provided

they

do not

lose

case.

figures trends. for

a significant failure provided This last load

portion

of their

howeer, 9 shows

qeneral

(n = 1.0 to 0.75) , and (close to zero) . might apply load in

in particular levels

P is small condition than

higher

correlation

the bene-

fits of low degrees i.e. n=2t06, n is high. however,


and with

of redundancy,

fatigue

rather

extreme

are not present The benefits with


lower

situations.

unless

increase,
(Fig. 10)

higher

B we

A similar, measure ~~ given

but more

direct

COV, s , as

of rednda. cy is the _probability of system fe. ilure This

have when case

we compare

the North

Sea

to the Gulf ideal

of Mexico

case.

Thj s

any

first member cf the to the

failuxe.

simple does

system

representation capture the small

is the ratio probability

system

failure

not of course

first merniJer failratios are

but potentially in mean exist

important

difference ratio that

ure probahiliti{. 0.12 (X-1-CP) .md

These 0.003

resistance/load bents,

(x-1-NS) .

between

nor the effects by vertical note For both t.h.atthe on the as the redundancy system Soad u,easres we

of initial loads .

forces

induced

redundancy ckarcictcris. tics itself . The

depends Seeral probabilistj. c meascres ca be as well

structure

of tbe reaur.iar, y of a system c suggested. For example

the difference

redundancy measures for zre s~nmarized in Table

the Iv.

structure

207

that Table IV Redundancy meas=r<s

the ratio

of frost-likely to rcostmay

failure-path

probability

-
Case <edunclancy * neasure 1 X-I-GM X-1-NS K-1-GM K-1-NS
0.12 0.16 0.05 0.05 0.12 0.003 0.7 0.4

like ly-first-m,c,.:k>c potability r be a good Redundancy * measure 2 ratio;

estin.zte of the preferred

.ystcrr.probabilj. ty to any

first -member-f ?.i lure proh F.bility. For the x- I-G1! c~se tb.e ratio of nlostis

likely-failure-path 0.14, which is very but mole ratio

to NLTF-nember close to the to

preferred, calculate

difficult

above.

A final

note:

one

should

fix? a

redundar,cy definition

that protects highly likely lead to defini-

Normalized

safety

index

difference.

ag~.i~st cases first member

where

** SYstem tional ure. (See text

failure to any

probability first

condi-

failures paths.

do not These

inember fail-

1~.kely failure tions

.11 require

further

thought.

for definitions) The semi-brittle used drop compression a 60% force this Fig. 5, full

The bility of any

first member above,

failure

proba-

member

model

assumes

used

is the probability failure. larger This the

instantaneous following

in member Comparixg

first-metier will be

failure.

probability

than member

load deflection with experimental

behaviour,

most-likely-to-fail-first probability, because

results , the

it represents of all It what

(60

%) capacity reduction requires

the probability possible has

cf the union

axial member shortening of approximately 5 times tbe failure deformation. For the X-braced structure, however, this deformation of a compression member was fcund to be less than twc times tbe failure deformation prior to the corresponding tension member failure. At this level of deformation, the remaining force in the compression member is in the order of 8G tc 90 % of capacity. The semi-brittle compression member
may thus be quite conservative A closer safety with the X-braced system. model for (but index

first member

failures.

the benefit be called

of representing system

might

complexity,

the more stressed

the number members will the

of highly larger this to the probaef fechere reduce

probability most likely

be relative failure hand

members

bility. tive

On the other

high

correlation

as is induced will

by high

load variability, between failure

the difference first-ner,ber

these

two

measures.

In this of the union member

study

the probabilityfirst-

somewhat may

unconservative) associated

of all possible

be that

a per-

failures

and the probability failure

fectly member. case, likely 4.40


to

ductile For the

(n = 1.01 compression in the GM of the most from first to

of the unict? of all possible paths are found not

this value, index

to be greatly associated with

safety

different most

from those first

failure 4.75,

path

iRcre*sed the

likely

member

arid most to the is

viz-a-viz

likely high

feil,~re path., owing

member

failure.

(4.75 c~rrespords

load COV.

The

implication

a probability

of 1 x 10-6) . This

hiqher

6S implies

a higher

re?,unciancy

chose

the dimensions
the uniLy were

of th, K- bxaces c
with respect Zur the

,Ceasure . The above 0.03. reduces

conditional

prehabil ity case to

so tlat
to two axial

check

fur the X-1-GM

loads

rratchcd

framicq

alternatives to the

wt.en API bad desig,, chose,,

desiqned

according XI the we

We conclude in the large

that COV

for Xbramss , even X-I-GM-case , with

quic?elines. to match in

insteaa ity leel one

total

ct.ecks , tbe would order have cf

tbe preferred

reclunc?ancy measure , probability any of

drop been

reliability by Ne

i.e. , the conditior.al system failure given

reduced

almost of

first member

magnitude. hae teria, aimec? chosen for at Cf

could

course matching c~uld

also crihave

failure, probably

is between less than

0.03 to 0.12 and 10-1.

some

other we

instance

r,atchinq the The in two

.th, reliability e Struct.c. al s, st~n,~ does what

K-braced the Kbraced structural not

Vertical

Bents.

In

Ieels directly. however

structure , the predicted is i,. general compared to the

criteriz

chosen represent

behavionr

principle of being

significantly case. and

changed

is and

thought

more

practical

the X-braced uniformity above there

1 particular discussed Therefore

realistic.

symmetry present.

is still is little

Given criteria the

this

API

based for the

matching resulting can iof axial

or o post-firstsystem capacity. The

reason

member-failure deterministic fails upon

difference partly tally. load axial for as be

in

reliability

level

semi-brittle of the Further,

structure first com-

explained

determinist in of given terms mean in in are the

failure

Utilizations ~ percentage are

pression failure

failure. deformation

tbe post

value Table the the two multiload V

of the compresare large repcOm-

capacity the critical Also

sion metnbe.s in the K-bays enough to make the

member qiven on

semi-brittle

structures. plication necessary The large thus

re.setation pared

less

conservative

factors to reach

wave

to the X case.

The q = 1.0

100 in

% utilization. this index factor is due cf

apprOXimatiOn unconservative results

would be ~ch more for these K-bays

difference the

The

and

reliability larger to used

of the piobabil istic structure 111 arc

analyses swnrna-

to Lt,e relatively
the x-brace capacity as

increase reach in the 1) the best

of the K-braced rized NS-case in Table

for the GM and and K- I-}:S large drop,

estmate bility also load of

means analysis

reliaand dead.

(as case

K-1GM
the

(see effect

Table of

respectively)

. Note

due

to

the

two orders reliability when

of maqnitude, level

in absclr.te system from the

reducing X-brace loads. of

tbe to

available be utilized is

capacity by a ~oll=e_ w?s.kness codes, that

for this

the

compared

to the results

~atexal quence Of

fhe latter obsered

X-braced

strctm-e.

a broadly

allc?w,able lead in to

.tres S based
much lower

Referring istically defined

back

to

the

determin-

they

overload that are

reserve be

strength

marqins stressed 10ad.5 events. effect we ratio

bracing under waves,

members extreme bir.ts noted larger Iie

fact(lr , F.EF , i.his would

2.3 f~r

only EGck, as It

lateral cr that if 10Z< Seisn!ic this the is

the K-braces

strurture , a si$,,lficat to the X-braced


c CC,r,e. S Firstly

re+ucticm> compared
REF of 3.3. 0?

shculd

be een to

Tkie. C3T-CF is se,.,eral factors.

we,,.ld Le of dead

que,lcc

load

209

!.,.r:zeasecf. One that K-braces ran-typical the structure

should analyzed

keep

in n,inci are in As a

.J!ie rx:de Ii iled is formed


The thus fcmr than after

as a truss , a mechanibm
first-member K-braced failure. is vit], system a

herein

as the were

story

heights

?Oux-bent more

structure

not changed. of the Kof the X-

directly ideal

comparable parallel the

conseqqen. e the batter L}:zccs nre twice braces. that

ccmpofient the X-braced failure

, as in

latter members in a

requirtis

both

bent, Table V Utilization critical in vertical factors bents. for in the

i .e. , a total fc~r-beut This

of eight

r.er,bers a

structure was however

to form taken

X-brace

and K-brace

mechanism.

into account when

in this

idealization as a

the x-brace .

was nmdelled

component

Utilization Factor

Multip.

factor sures

The proposed for the bents

redundancy with

neaK-braced in had of

on Fit to 100 Utilization

structure are

vertical Table

also given

Iv. rbe X-braced system

structure

T
-BRACE -B8ACE
ideal parallel

a conditional 0.27

probability even

4.36
3.03

failure high COV

of about case.

10-1,

in the this prob-

In contrast

0.33

ability

is O. 7 for the K-braced. ically would we concluded fai 1 when that

Determinist the K-brace member

a single

failed, might

whereas

the X-braced more

Repeating

the system,

analog as

to was

an done

system load

carry

a few percent on the

(O to 5 %, depending of the

for the X-braces,

the ultimate

capa-

effectiveness ness

remaining member

stiff-

city of a K-brace component is: 2(0.8) t.c. = 1.6 t.c. (recall the larger librium k~ fr ratio) . To maintain after failure occurs force equi-

in the tension

in avoid-

ing the drop pressive

to n = O. 4 of the comat failure) . System shows the X-

force

in the in the

reliability braced measure better versus observed

analysis to have

compress ion member, tension tbe post brace t.c. member must

tbe also

system about than about

a redundancy o f magnitude (about (!. 1

decrease, of the 0.4 x 1.6 in

an order

failure

capacity

that

of the K

system = 0.64

is therefore t.c. . This capacity

0.7) . It is been study of ideal paral-

results

in the e.g.

postfailure The effective

factor, n = 0.4.
remains

lel system, 10, here)

Guenard

[see Fig. (6) ,

correlation

and

Stahl

and Geyer

abcut small mostis not

90 %. Referring reduction in found

to system in In the fact,

Fig. versus

9,

the

that

the absolute can

level the this

of the reliaimpact of

bility

influence

likely-first-member-failure analysis with


system tha.r.th.lit such sur-

rednndanc y. To test

issue we system with

probabilities surprising.
e low n level prised

h?.ve ar,alyzed a K-braced bracing the capacities

increases

to give

one might

not be

same most- likely -first-nember probability as the X-brace. remained level at

to find a four-bent greater

failure

faill,re probability individual not iF.q that

rl,e redundancy 0.7. At the

measure COV

bent.

It is also bent,

lower

the Xis

in a K-braced

brccecl systems

redundancy

measure

210

LWO orders

of

magnitude

better.

(Note
to a

An

important

effect

c.f c%n,zqe is

a~ciin the dependence Ckncy measure on the

o f the ~edur,load Cov] Con.

tension

r,enber in a X-brace length

that cent

the buckling compression

of the adjain-

cluchng,

the K-braced

structl-e F.?. c redundancy case, effect

nia.,berwill

O1.l a sn.al system Y 1 even cular for the when low CGV

crease, Also

i .e. , its capacity will

decreases. increase

in. partiin

its deformation the

compared

to the effect

causing

semi-brittle

model

(n = in

the X-braced

structure.

O .4) to beccm,e less the following, and without capacity, both in a

conservative. options, i .e. ,

X-braces Damaged

Versus Structure

K-braces

with

reduceti compression are ciiscussed.

member

General In ordez to aSSe SS the u impact of damage on system reliability, members caused fatigue tic the jacket as analysed may with be

Damaged = thought The

X-braced member

Vertical damage for the

single

to be most

critical

removed.

The damage

structure, member tension

is damage

to tbe tension X-bent. With this

by an accidental or defects. Lloyd

loading,

in the outer member

In a dete~inisK Clawson (2) to

damaged, bent

the compresis the one The when the

framework

sion member most results likely

in this to fail

measure sustain dual

a structure *s ability a member failure

first.

by its resiResidual

found

for the case is assumed support

strength,

defining

ciamzged member to provide compression


at

to be able to this

Resistance

Factor

(iiIF) as the ratio


load at collaps load

lateral member

of environmental
damaged coil.aps tic to the

are denoted 111. The failure safety in the This

environmental Prom will to in

X-D-GN (A) in Table index first representing (or next) than

nndamaged. we

a probabilisassess damage system the by com-

viewpoint

member

is 3.15.

system paring psobabil to that

sensitivity the ity of increase in the the

is lower

that case

of the

same member to

failure

in the intact the increased

( B = 4.28) , due of the

damaged

condition

utilization

intact

structure.

compression i; 2.3x1O (under the sion member failure Structure bility been

member.

!Cbe probability under storm waves that

of syst@_~ilure
Only indices lifetime system safety

( 6 = 3.50) . Recall same semi-brittle the assumptions)

have

been

calculated

for the

compressystem

damaged such

structm-e.

In a real project could be useful the effect in of

information

probability

of the intact The probafailure has

the case of assessing undiscovered in connection necessity of known obtain give damaged, with

is 1. 3X10-5. induced

for instance, the

of storm

ealcating

increased. by a factor

of about

of inspection. damage, one

In the event however,

20 hy the presence daraged measwre respect fail~xe. system failure member. of the

of a previously is a with

should

This change system

the annual

6 1s as they would i determintc the

robustness

useful

information

to So,e exogenously (A statically have of unity given

c&use~

ing i f repair following diate

can be postponed season,

determinate cf a damage;

summer

or if irmne -

would

a probability

action

is required.

perfectly,, robust was not effected would have


by

system,
the

tir,e that.
of a

loss

member

no increase

in

211

systm

robstess

has

deczease<:

,;i, th onc

a f?rtor If the ciamaged tension is not able sion member buckling doubied, to support in the member member to the the

tvc cumpared system. failure

to the only With

damaged system

respect of

the ccm~res the is

probability is

same bent, member

intact

structnre , this by a factor analysis

length

of this

ihcreased The

of 55. that have the suffi-

and t}xe compression is reduced system

capacity

showed still
almost

of the member nately

by ap~roxi probability (ce. e i implying a

horizo],tai X-brace cient


keep

20 %. The

capacity
the failure bents but to the

to

certainly the

of failure x-D-GM(B) robustness

is 3 .4X10-4 in Table factor 111)

path and at

withiu the

vertical level, start

ti.amaged X-brace do

of 25.

horizontal potentially Further as members tbe it

become

susceptiwas in the

It is interesting by comparing ture cased, these that

to confirm, strucas first

ble

to

failure. that bents,

two damaged

observed vertical planes side

fail

redundancy versus

remaining X-planes, unfailed broadvertical utilizations

measured member being

here failure)

(system

(horizontal and

is related

to there

diagonals have

non-uniformity

in relative (or bents) . system failure capacis

X-brace)

comparable

utilizations In these

of nenbers the of

and

B s.

two cases,

ity and probability about case the the same, but

Damaged Vertical structure removing sion

Structure

-K-braced

in the member

second in the higher

Bents. damage

In the K-braced was simulated by

remain:nq bent

damaged loaded

is initially

a member

loaded The

in compresresults from

relative

to its capacity first-member fail-

in an end bent. are

implying ure

a higher

the analysis 111 case

summarized

in Table

probability.

K-D-GM.

In the event bay being shear acting damaged

of a full end-bent (no load bearing/ loadlshear have to be bents above. In unlike have

One

should

keep

in mind

that would

the X-brace, active

which

still

capacity) on that

, the wave bent will

one

member

left,

the bay

in the one-member-damaged structure Comparing ure in the txausfers probability intact

K-braced load. fail-

transferred ia this duces next

to the remaining X-braces

no shear

the horizontal structure a large

cf system

the unsymrnetry shear force

intro-

(GM) versus the system

the proba-

in the bent

damaged bility 1.3x10-3 This

structure failure intact

tc the damage.

Tbe probability member

increased to about

from about 5.3x10-2.

of failure in this high, The

of the compression is therefore


III case

bent

relatively
X-D-GM(C) .

is an increase failure

in wave-caused of about damaqed this is of as that

see Table

system 40 due

probability

remaining still

part bas

of the some

structure, for

to the exogenously r.wr,ber.Note of system

however, load

capacity this

(removed) measure

that

increase. displays

Therefore strong

dzmase

robustness

system

redundancy failure

the: szme orc?er of magnitude of tbe X-braced system.

vi th respect

to wave

induced

as r.easured by the proh?bility

system

failure in

conditional

to failure

the f4LTF r.ember probabi 1ity. The

net

212

Role

of Horizontal system

Bracinq

in the

frames, replaced

the horizc,uizl by diagonals

X-traces

were

~rctural

s:icqle diagonal , as not

membei-s. The I.emher dimensions f The jacket ers$ horizontal sized bracing using in the desiqn cocie givex by the aboe rules, were

has keen

changz?. this

T>.C ~eliakiiity structure fa~lure

zr.cl>zi$ c.f that not dnd of failZarna.ged of

prerogatives

r<. ti-.erthan

intact

revealed nc, de was

criteria, ly unloaded drinq real

?s this bracilig is basicalin the intzct structure In a

the dozcir, ant influenced that the

by this systen

reframing,

its operational

phase.

probability changed. agsin The

structure loads

the on-baxqe VOUIC? have

transpor-

ure was case

hardly

tation

determined membas

investigated

consisted

the dimension bers; only this been

of soae of these however,

removal end-bent Even

of a tension

me fnher in the leel bay.

structure, designed

at the critical case, appeared

for the opexaFor the structure

for this mode

the most

likely

tienal with

conditicn.

failure

in the vertical failure in the

vectical braces

x-brace. , the horiwere sized according

X{ s. However, mode consisting

a erticaln of failure

zontal
to:

horizontal from

diagonals
bent

frar.ing inwards
appeared as

the damaged
most likely.

Minimum

wall

thickness , cm (3/8 inch) ratio,

next

t = 0.95 Diameter

Concluding, structure framing diagonals showed consisting was

the that of

analysis

of

tbe

tc thickness D/t < 60.

a horizontal only simple

sufficient

to fulfil of this is subject is, the

Furthermore ur.aticm (x,s) was arbitrarily

tbe bx.acing con figapparently This rather

the Structr.al requirements frami Eq when to a broadside horizontal the structure that

chosen.

sizing

wave,

apprOach resulted in a rather stiff horizontal framing, with k 9./r= 35 and thus pression only versus 10 % reduction tension i com-

framing

did

not contribute ~ystem this lower

to a significantly prchability probability of

higher

failure

in fact been

strength.

would

not have were

if the horizontals Analysis ture wave, braces subject pr.xed of the intact struc strong.

infinitely hand, the

If, on the other

tu a south that

(broadside)

capacity of the horizontal further , its members would


the reliability (Note that ture. this

framing
control struc-

the horizontal their mission and the by leel -

system was reduced significantly


of the damaged

fulfilled

efficiently ling out tical

transferring

loads between Een

four er -

conclusiGr, xzy wse approach

bents.

iR the clamaged and

not be valid directions) .

for other

post-failure

cases , the horizontals

appeared to be strcnq ermugb both with respect to carry .nrj forces and stiff nesswise. hae surplub In fact they capacity seemed to

CONCLUSICNS

in all cases.

Tb.e

example

structure cases

has bem incluai,lg

analyzed In order the to ufiderstan< better sr, insuffiti

for seeral

X and K braces,

high and low load


these arm lysis a

limit bc,kv~ec,, excess strength

variability and intact ver scs damaged states. Through

cient

c? the t.c, rizontal

213

,iL-Gctural SiS Scme !,T metlmd r-suits

system has been

reliability demonstrated. in

an&ly -

are

summarized

fai,le

From

these

results

it can he

cor, eluded ix the K-braced with

that

the X-br.acea framing is superior to the both level wixh tbe

khstr,ess of the
respect increase to damage ir, system damage With

StICt

S~

structure

is mea :L. d b>re. failure proba --

as designed.

ibis kolds safety

respect

t.o ahsclute wheu both

bility

following member.

(remoa.1) cf this raeas=re was fcmnc?

and redundancy, systems

framing to API

a critical

are desig~ed

according

no significant between words the two

difference systems. the

and compared design unity

on the basis checks.

of axial

:n other X-brace lcc-I system :or the

Detenninisticalfactor REF for

although

individual safety

lY, tbe reselve for the K-brace the X-brace. increase failure The

strength

has a higher

absolute

is 70 % of that correspon6inq probability than

and red~ndancy, behaviour

the overall

is not dissimilar

in cystem is larqer

.f of

two systev.s.

cr.e order

magnitude.

The interesting

study

has revealeti some ith respect to

aspects

The is twofold, tively

reason

for this due

difference

redundancy

and robustness state the

measures . structure bas

firstly

to the relain capacity best of

In its intact high tion. uniformity This

larger

incresse

in ner,ber utilizain a low redundano f a member in utilizaand

the X-brace mate values

when

comparing

esti-

results Damage

tc design

values,

and

cy measure. results tion

secondly cal loads

indirectly

due

to the verticarried increase

irL r.on-uniformity

(dead load)

being

in the remaining

structure,

by the X s and the in lateral overload

resultant margin.

as a consequeucc

hiqher

redundancy.

The Tbe X-braced larger system structure showed than a cwnsistency structure,

result

is au apparmt a well

~.n-

in that

designed

redundancy This

effect

i .e. , well

balanceri, ir.plies LmplI, that


lGW

the K-braced. evident in the

is in particular

highly--utilized-members, redundancy-. redundancy quality? measure of Eoes is not Does this

low load variability of member xore

case when resistances

the variability become With

as a desirable that the

relatively the

it imply here

sj. r. icant. q if re?=ndaucy system member x-braced

suggested

introduced

is inapfor

measure

of probability

propriate,

.althcugb it works

failure fa~lure,

conditional

or. first of the ; times

the redundancy is minimum even case.

statically determinant e system r>? OLdce. ic ~Lean that ,,robstness is what. we are really more interested like a major

structure

the.f of the K-braced, i-ich load variability

in the

ir. i.e. , we wouiti not , reduction the in system

rel j.ability if damaged? dis-

struccuzt issues

is lccaily should

study showed that the sj.~ie. to the mOdel redundar.ci is ,:., ling of member iour. Fuzkher post-failure the redundancy bek, avwill

The

These

be further

in tt,e ;Lgbt cus S!?cl measuring reliability

of tl,e improved Of SYSteM

capabil jt?e~ allaiysls.

214

Table

VI

Summary

of st~dy

xesults.

G14 = Gulf h-S = North

of lNexico ihigh load variability) Sea (low load arj. ability)

BRAcE TYPE

PROB.

OF F>.l.l,C. F,E

IWGUNDiUICY MEhSUhE

pROB.

OF FAILURE DAMAGES

INTACT ~;~ -- __ ___ .

GM.

G&i

Ns

G!l

1.3X1O -5 1 .1X1 O-6

0.1

0.003

3.4

X10-4

1.3 X1 O-3

3.9 X10-5

0.7

0.4

5.3 XI0-2

It is worth real example platform, structure,

noting

that

More

sophisticated

as opposed would

to the

model ling with ing. Capabilities gent of

respect

3
NESS MEASUR G14 25 40 load patter to scalof herein was Industry System <uring the Nordal , of the availzbl e also hich in tibe K.om_ like to to scott r!a.xccn ,

ROBUST

be more require-

non-cniform ments

due to design prior

,,more intelligeneration

from phases

to the Althowgh in

automatic trees.

i,,place operational unintentional higher system this

phase. would

failure

result

post-first-member-failure capacities and a more redu-

ACKNOWLEDGEMENT

dant bchaviour. The member nique, system proved which replacement techProject

The initiated

study

reported

through

the Joint

on Offshore chaired

Structural by Bernhard principal C .Allin

is the basis

for the has in

Reliability, Stahl , AMoco investigator

reliability

analysis,

and with Professor

to be an efficient

tool

comparj.nq alternative systems and their

structural or with Scope

Cornell . It was stay at Stanford in 1986.

performed of Earald

redundancy sreas

robustness

Hoeverr

Statoil

for improvement

in the tech~,iqe , as
The structural

~plemented in the computer program FAILUR, have been identified.

model was made

example for the

structure study tbe

by EXXOn, structure

The most

important

are:

redesigned braced

case . Cheron time to the

provided

E!Pre detailed frealistic

mechani-

puter mo~t
750

s,tdy, although on a Stan forO. VAX

cal rnodelling of post-failure compressi cn member Member axial behaviour. ccmlhined

runs m+re
Canpxter.

made

beh.aviour under

an? IXXW?nt. Icadj rm.

The ekpxess

authors

ould

their

appreciation Hugh

Mancindale

at Chercm,

215

6.

B. Stahl matc

and J.F.

CCye Ir ,,lti U of

Strel!gti; RelikbiiltyLeg

Tension

Pla. tfc, rm r+!ition YS S

telns,, 17th ?,nnual Cff,hc,rc , Technclm?y !. Y. E-. Guenard: ~llcatiOn Of system Reliabi]it}, Analysis to Offshore Structures, John A. Blume Earthquake Enq. Conference, ?roceed 4S57,

irrgs, liousto~, OTC paper My, 1985.

Center

, Stanford 1984.

Univ. , Report

No .71, Nov.

2.

J.R.

Lloyd

and W.C.

Clawson: strength platof

Reserve pile

and

residual of fsbore

founded,

fxlls,, The , Inspection Marine

Role

of Desiqn, in

and Redundancy

Structural

Reliability. Academy 19S4.

Proceedings, Press,

National DC,

Washington

3.

A. Karamchandani, System Reliability

Structural Analysis

Methods, Department of Civil Stanford University, Engineering,


Draft report to the Offshore joint industry project Systems Structural Feb. 1987.

Reliability,

4.

F. ary

Moses: Load

Development and Resistance


for Fixed API-PRAC Inst. Ohio,

of

PreliminDesign

Document
Platforms< 85-22,

Offshore Project of Technology,

Case

Cleveland,

January

1981.

5.

tf. D. AndersGr,,
J.R. dure Lloyd, for Fixed

!.!.1:.

Silbert

and
PKOCe-

Reliability Offshore

Plat-

forms , Journal ASCE of Structural Division, NOV.

VG1 . 108, No.

ST1l,

1982,

pp.

2517-2538.

216

Você também pode gostar